Bcps Study Design And Statistics

Front 1

What is a description of cohort study design?

Back 1

A cohort study starts with two groups without the disease, separates them into groups based on the exposure of interest (smokers vs non-smokers), and follows to see how many develop the disease.

Front 2

What is a description of case-control study design?

Back 2

Select a group of cases with the disease, a group of controls without the disease, and see who was exposed to the proposed cause.

Front 3

Is a cohort study retrospective or prospective?

Back 3

Prospective

Front 4

Is a case control study retrospective or prospective?

Back 4

Retrospective

Front 5

What is nominal data?

Back 5

Data with incomparable values (gender, yes or no)

Front 6

What is ordinal data?

Back 6

Data that can be ordered but without degrees of difference (most to least)

Front 7

What is continuous data?

Back 7

Data that is numerical and allows for difference (temperature or blood pressure). Interval and ratio are examples.

Front 8

What is paired data?

Back 8

Data that arises from the same individual at different points in time.

Front 9

What test(s) for nominal data with two independent samples?

Back 9

Chi squared, Fisher’s exact

Front 10

When is Fisher’s exact test used?

Back 10

With nominal data from two independent samples that are too small to use Chi squared.

Front 11

When is Chi squared used?

Back 11

With nominal data with independent samples. For more samples remember to apply Bonferroni.

Front 12

When is Bonferroni adjustment used?

Back 12

For Chi squared with more than two comparisons

Front 13

What test(s) for nominal paired data?

Back 13

McNemar’s test

Front 14

When is McNemar’s test used?

Back 14

With nominal paired data

Front 15

What test(s) for nominal data with three or more independent samples?

Back 15

Chi squared

Front 16

What test(s) for ordinal data with two independent samples?

Back 16

Mann-Whitney-Wilcoxon U-test, Wilcoxon Rank Sum (either), Kolmogorov–Smirnov

Front 17

When is Mann-Whitney-Wilcoxon (U-test) used?

Back 17

It is based on the median of ordinal data or non-parametric continuous data with two independent samples.

Front 18

What test(s) for parametric independent continuous data with one or two groups?

Back 18

T-test

Front 19

When is t-test used?

Back 19

Continuous parametric independent or paired data with up to two groups

Front 20

What test(s) for ordinal paired data?

Back 20

Wilcoxon signed rank test, or sign test

Front 21

When is Wilcoxon signed rank test used?

Back 21

With ordinal paired data

Front 22

When is the sign test used?

Back 22

With ordinal paired data, a less demanding Wilcoxon signed rank test

Front 23

What test(s) for ordinal data with three or more independent samples?

Back 23

Kruskal-Wallis

Front 24

What test(s) for continuous parametric data with three or more independent samples?

Back 24

ANOVA

Front 25

What test(s) for continuous parametric data with three or more independent samples and confounders?

Back 25

ANCOVA

Front 26

When is Kruskal-Wallis used?

Back 26

With non-parametric ordinal or continuous data with three or more unmatched groups. It is an ANOVA variation.

Front 27

When is ANOVA used?

Back 27

With ordinal or continuous parametric data of multiple samples to tell only if there is a difference anywhere.

Front 28

When is Friedman test used?

Back 28

With continuous non-parametric or ordinal data in three or more paired groups.

Front 29

When is Cochran–Mantel–Haenszel used?

Back 29

With nominal paired data in three or more groups.

Front 30

What test(s) for ordinal paired data with multiple attempts?

Back 30

Friedman test

Front 31

What test(s) for nominal paired data with multiple attempts?

Back 31

Cochran–Mantel–Haenszel

Front 32

What test(s) for normality (parametricness) of continuous data?

Back 32

Shapiro–Wilk, Kolmogorov–Smirnov (Lilliefor), or Anderson–Darling tests. Differences are hopefully beyond BCPS.

Front 33

What test(s) for correlation of parametric data?

Back 33

Pearson

Front 34

What test(s) for correlation of non-parametric data?

Back 34

Spearman

Front 35

What correlation coefficient (r) value(s) shows greater association between variables? What test is this associated with?

Back 35

Close to -1 or 1. Ranges -1 to 1.

Spearman test for correlation of nonparametric data.

Front 36

What coefficient of determination (r2) value(s) shows the regression line fits the data? What test is this associated with? What would a result of 0.8 mean?

Back 36

Close to 1. Ranges 0 to 1.

Pearson test for correlation of parametric data.

80% of the variability of Y is explained by the variability of X

Front 37

How does power relate to the confidence interval?

Back 37

Low power will have a wide confidence interval, good power will have narrow confidence intervals.

Front 38

What is an acceptable power level?

Back 38

20%

Front 39

What is the difference in definitions between correlation and regression?

Back 39

Regression is predictive, correlation is not.

Front 40

What is the formula for standard deviation?

Back 40

Square root of variance. Variance is squares of difference from the mean summed, then divided by n.

Front 41

What is a definition of the standard deviation result?

Back 41

Values that amount above or below the mean are one standard deviation away. Requires normal distribution.

Front 42

How much of a normally distributed sample falls within 1 and 2 standard deviations from the mean?

Back 42

68% and 95%

Front 43

What is a definition of Standard error of the mean (SEM)?

Back 43

The standard deviation of sample means over all possible samples drawn from the population.

Front 44

What is the formula for standard error of the mean (SEM)?

Back 44

SEM = SD / square root of n

Front 45

What is the layout of a risk table?

Back 45

Columns left to right are whether disease is truly present or absent. Rows are whether subject is exposed to suspected risk factor.

Front 46

What is the layout of a contingency table?

Back 46

Columns left to right are whether disease is truly present or absent. Rows are whether test resulted positive or negative. (a,b,c,d = TP, FP, FN, TN)

Front 47

What are the formulas for Experimental event rate (EER) and Control event rate (CER)?

Back 47

EER = a/a+b, CER = c/c+d

Front 48

What is a definition of absolute risk reduction (ARR)?

Back 48

The simple difference in incidence between active and control groups.

Front 49

What is the formula for absolute risk reduction (ARR)?

Back 49

ARR = EER – CER or a/a+b - c/c+d

Front 50

What is the formula for number needed to treat (NNT)?

Back 50

NNT = 1 / ARR

Front 51

In what studies may relative risk be used?

Back 51

Prospective studies, like cohort or experimental

Front 52

In what studies may odds ratio be used?

Back 52

Retrospective studies, like case-control or cross-sectional

Front 53

Whatis the formula for relative risk (RR)?

Back 53

RR = EER / CER or (a / a+b) / (c / c+d)

Front 54

What is the formula for odds ratio (OR)?

Back 54

OR = (a/b) / (c/d)

Front 55

What is a definition of relative risk reduction?

Back 55

The ratio of risk difference between exposed and unexposed groups.

Front 56

Whatis the formula for relative risk reduction (RRR)?

Back 56

RRR = (EER − CER) / CER or 1 - RR

Front 57

What is a Hawthorne effect?

Back 57

When behavior changes due to being aware of being observed.

Front 58

What is a definition of any given null hypothesis?

Back 58

There is no difference between the groups.

Front 59

What is a definition for Type I Error and how does it relate to significance?

Back 59

Alpha. Rejecting the null hypothesis when it is true. By definition, can only be considered if a statistical difference is found.

Front 60

What is a definition for Type II Error and how does it relate to significance?

Back 60

Beta. Accepting the null hypothesis when it is false. Should be considered when a result is NOT significant.

Front 61

What is a definition for sensitivity?

Back 61

True positive rate. The proportion of positives that are correctly identified as such. Avoidance of false negatives.

Front 62

What is the formula for sensitivity?

Back 62

Sensitivity = TP / (TP + FN) also a / a+c

Front 63

What is a definition for specificity?

Back 63

True negative rate. The proportion of negatives that are correctly identified as such. Avoidance of false positives.

Front 64

What is the formula for specificity?

Back 64

Specificity = TN / (TN + FP) also d / d + b

Front 65

What is a definition for Alpha?

Back 65

False positive rate. Rejecting the null hypothesis when it is true. Type I error. p = Alpha

Front 66

What is a formula for Alpha?

Back 66

1 – specificity = 1 - TN / (TN + FP)

Front 67

What is a definition for Beta?

Back 67

False negative rate. Accepting the null hypothesis when it is false. Type II error. Power = 1 - Beta

Front 68

What is a formula for Beta?

Back 68

1 – sensitivity = 1 - TP / (TP + FN)

Front 69

What is a definition for power?

Back 69

Power is sensitivity; the proportion of positives that are correctly identified as such.

Front 70

What is a formula for power?

Back 70

TP / (TP + FN). Equivalent to sensitivity, or 1 - Beta.

Front 71

What is a definition of p-value?

Back 71

The probability of obtaining a result equal to or more extreme if the null hypothesis were true. The type I error rate is at most alpha.

Front 72

What is a formula for p-value?

Back 72

Data, a null hypothesis, and alpha level, a t- z- or f- test and a computer or a table.

Front 73

What values of a confidence interval show statistical significance in prospective trials?

Back 73

Any that does not include 0.

Front 74

What is a definition of Kappa?

Back 74

Measures inter-rater agreement for qualitative items that takes into account the agreement occurring by chance.

Front 75

What is the range and interpretation of a Kappa value?

Back 75

-1 to 1, -1 is a systematic disagreement, 0 is chance, and 1 is perfect agreement. Higher is better, meaning the raters are very similar. Not learning formula.

Front 76

What test(s) for non-parametric ordinal or continuous data with one or two groups?

Back 76

Mann-Whitney-Wilcoxon (U-test), Kolmogorov–Smirnov

Front 77

When is Wilcoxon Rank Sum used?

Back 77

Ordinal data with two independent samples.

Front 78

What test(s) for parametric paired continuous data with one or two groups?

Back 78

T-test

Front 79

What test(s) for continuous non-parametric data with three or more independent samples?

Back 79

Kruskal-Wallis

Front 80

What test(s) for continuous non-parametric data in three or more paired groups?

Back 80

Friedman

Front 81

What are the numerical conditions for using Chi Squared instead of Fisher's Exact?

Back 81

where n is at least 20 and frequency per cell at least 5. For more samples, each cell must have at least 1 and >20% at least 5.

Front 82

What is a definition for incidence?

Back 82

The number of new cases within a population in a time period.

Front 83

What is a definition for prevalence?

Back 83

The number of cases of a disease in a population at a given time.

Front 84

What is the equation for positive predictive value?

Back 84

A / (A + B)

Front 85

What is a definition of positive predictive value?

Back 85

The proportion of positive results that are true positive results.

Front 86

What is the equation for negative predictive value?

Back 86

C / (C + D)

Front 87

What is a definition of negative predictive value?

Back 87

The proportion of negative results that are true negative results.

Front 88

How does prevalence relate to sensitivity, specificity, and positive and negative predictive values?

Back 88

Prevalence can be used to calculate predictive values, but has no relationship with sensitivity or specificity.

Front 89

What is the definition of per protocol?

Back 89

Once randomized, only patients who adhere to protocol are reported in outcomes.

Front 90

What is the definition of intention to treat?

Back 90

All patients randomized are reported at the conclusion of the trial regardless of adherence or outcome.

Front 91

How are delta and confidence intervals used to assess the null hypothesis in non-inferiority trials?

Back 91

Delta is the largest clinically acceptable difference; the comparison treatment must be "not much worse" than the standard treatment. The confidence interval may pass below the effect of the standard treatment, but must not pass below delta or the comparison drug is not non-inferior.

Front 92

What is the general p-value and confidence interval for a one-sided test?

Back 92

p-value: .025

confidence interval: 97.5%

Front 93

Is per protocol or intention to treat more appropriate for a non-inferiority trial?

Back 93

per protocol is preferred as it is more difficult to show non-inferiority in this way. Both would be nice.

Front 94

What tests for heterogeneity in meta-analysis? What statistic is used? What does it mean to reject the null?

Back 94

Cochrane Q.

Uses Chi Square and a p value.

If result is less than p value, then reject the null, meaning there is too much heterogeneity to pool results at face value.

Front 95

How can an over or under powered meta-analysis be studied for heterogeneity. What does the result mean?

Back 95

Cochrane Q will not work well. Use I^2. O% is best, meaning chance alone. >50% is high heterogeneity.

Front 96

What does a highly incohesive forest plot indicate?

Back 96

High heterogeneity between studies. There may be considerations to take with the study targets or methods.

Front 97

How is funnel a funnel plot made? What does it indicate if weighted to one side?

Back 97

y axis is study size, x axis is effect result. Publication bias.

Front 98

What is interval vs ratio data?

Back 98

Ratio data has a true 0 meaning no amount (weight), whereas interval data does not (temp in farenheit).